Nativ Puts Frontier Models on Your Mac, No Cloud

Prince Canuma, the person behind mlx-vlm, shipped Nativ, an app that downloads and runs open frontier models entirely on your Apple Silicon Mac. No account, no subscription, no cloud round-trip. It runs language, vision, video, code, and embedding models locally, and it's free, private, and open source. Because it's built straight on MLX-VLM and tuned for M-series unified memory and Metal, there are no wrappers or translation layers eating your performance.

The timing is the point. Canuma has a habit of shipping day-zero support the moment a new open model drops; mlx-vlm v0.4.3 landed the same day Google put out Gemma 4. So the moment Qwen 3.8, Gemma 4, Kimi K3, and the rest of this open-weights flood hit, there's a clean path to running them on the laptop already on your desk. Chat with an LLM, caption an image, summarize a video, autocomplete code, transcribe or generate speech, all offline.

This is the quiet counter-story to the trillion-dollar datacenter build-out. Every week the open models get better and the machine to run them is already sitting in front of you. For anyone who cares about privacy, cost, or just not sending their work to someone else's server, local-first stopped being a compromise. Get it at blaizzy.github.io/nativ.